fix old "Document" import (#3)

- fix old "Document" import (933edc57044ac89ba060526525a9bdfdb9b5acff)
- update reqs (49a2a7185d954271fb50dc7f9dcd7befd2d9b60c)
- download HF models during image build (5c02889da4e27203b934318b9deb5c23dd140f47)
- Pre-download Hugging Face models during build (06d3da632b935e3480a41c9cab85c18d42b89613)


Co-authored-by: Ara Yeroyan <Yeroyan@users.noreply.huggingface.co>

Dockerfile

@@ -15,6 +15,12 @@ COPY requirements.txt ./
 # Install Python dependencies
 RUN pip3 install --no-cache-dir -r requirements.txt
 
+# Pre-download Hugging Face models during build
+# This caches models in the Docker image for faster container startup
+COPY download_models.py ./
+COPY src/config/settings.yaml ./src/config/settings.yaml
+RUN python download_models.py
+
 # Copy all application files (excluding .dockerignore patterns)
 COPY . .
 

download_models.py

@@ -0,0 +1,54 @@
+"""
+Pre-download Hugging Face models during Docker image build.
+
+This script loads the models to trigger download and caching.
+"""
+import os
+import sys
+
+print("🔽 Downloading Hugging Face models during build...")
+
+# Model configurations from settings.yaml
+EMBEDDING_MODEL = "BAAI/bge-m3"
+RERANKER_MODEL = "BAAI/bge-reranker-v2-m3"
+
+try:
+    print(f"📦 Downloading embedding model: {EMBEDDING_MODEL}")
+    from langchain_community.embeddings import HuggingFaceEmbeddings
+    
+    # Load embedding model (will download if not cached)
+    embeddings = HuggingFaceEmbeddings(
+        model_name=EMBEDDING_MODEL,
+        model_kwargs={"device": "cpu"},  # Use CPU during build
+        encode_kwargs={"normalize_embeddings": True},
+        show_progress=True,
+    )
+    
+    # Trigger actual download by encoding a small text
+    test_text = "test"
+    _ = embeddings.embed_query(test_text)
+    print(f"✅ Embedding model downloaded: {EMBEDDING_MODEL}")
+
+except Exception as e:
+    print(f"⚠️ Warning: Could not download embedding model: {e}")
+    # Don't exit on error - allow build to continue (model will download at runtime)
+    pass
+
+try:
+    print(f"📦 Downloading reranker model: {RERANKER_MODEL}")
+    from sentence_transformers import CrossEncoder
+    
+    # Load reranker model (will download if not cached)
+    reranker = CrossEncoder(RERANKER_MODEL)
+    
+    # Trigger actual download by running inference
+    test_pairs = [("test query", "test document")]
+    _ = reranker.predict(test_pairs)
+    print(f"✅ Reranker model downloaded: {RERANKER_MODEL}")
+
+except Exception as e:
+    print(f"⚠️ Warning: Could not download reranker model: {e}")
+    # Don't exit on error - allow build to continue (model will download at runtime)
+    pass
+
+print("✅ All models downloaded and cached successfully!")

requirements.txt

@@ -1,9 +1,24 @@
+pydantic>=2.0.0
+torch>=2.0.0
+numpy>=1.24.0
+pandas>=2.0.0
+FlagEmbedding==1.3.5
+sentence-transformers>=2.2.2
+transformers>=4.35.0
 streamlit>=1.28.0
-langchain>=0.1.0
-langchain-core>=0.1.0
-langgraph>=0.0.20
+langchain==0.3.25
+langchain-community==0.3.24
+langchain-core==0.3.79
+langchain-huggingface==0.3.0
+langchain-mistralai==0.2.10
+langchain-ollama==0.3.3
+langchain-openai==0.3.23
+langchain-qdrant==0.2.0
+langchain-text-splitters==0.3.8
+langgraph==0.6.10
 qdrant-client>=1.7.0
 python-dotenv>=1.0.0
 openai>=1.0.0
-snowflake-connector-python>=4.0.0
-pydantic>=2.0.0
+pyyaml>=6.0
+tqdm>=4.65.0
+snowflake-connector-python>=4.0.0

src/pipeline.py

@@ -4,7 +4,11 @@ from pathlib import Path
 from dataclasses import dataclass
 from typing import Dict, Any, List, Optional
 
-from langchain.docstore.document import Document
+try:
+    from langchain.docstore.document import Document
+except ModuleNotFoundError as me:
+    print(me.__str__())
+    from langchain.schema import Document
 
 from .logging import log_error
 from .llm.adapters import LLMRegistry